A model for minimum ignition energy prediction of sugar dust clouds based on interactive orthogonal experiments and machine learning

This study applies the concept of orthogonal experimental design to explore the influence of ignition delay time, dispersal pressure, dust mass, and interaction effects between these factors on the minimum ignition energy (MIE) of sugar dust. Additionally, it aims to establish a comprehensive predictive model for the MIE of sugar dust considering the combined effects of multiple factors. Comparative analyses were conducted between multiple linear regression polynomial models, Back propagation neural network (BPNN), and Generalized regression neural network (GRNN) for data fitting and prediction accuracy. The results demonstrate that the MIE of sugar dust is significantly influenced by dispersal pressure, ignition delay time, and dust mass, while the interaction effects between these factors are not statistically significant. Artificial neural networks exhibit superior predictive performance compared to traditional multiple linear regression polynomial models when considering the combined effects of multiple factors. Specifically, the GRNN model outperforms the BPNN model, highlighting its high fitting accuracy with limited sample data.